* Wiesław Kielas <wieslaw.kie...@bluemedia.pl> [04.09.2013. @15:44:05 +0200]:
> Hi misc@,
> 
> I have a Dell PowerEdge M600 machine running OpenBSD 5.3 which causes
> frequent problems - once about every few days vlan interfaces stop
> working.
> 
> Ifconfig reports them being up the whole time, but when trying to ping
> anything in the given vlan, the ping fails (this also applies to the
> loopback address on the interface - it doesn't reply as well), as if no
> packets were transmitted on it. This applies to all packets on the
> interface, not just ICMP ones.
> 
> This doesn't seem to be a PF issue, since disabling it doesn't help, and
> non-vlan interfaces (like trunk0 with trunkproto failover) continue to
> work.
> 
> Putting the interface down and up, destroying and creating it or running
> /etc/netstart doesn't help - a reboot is required, after which the
> machine works correctly again for some time.
> 
> I checked the logs and dmesg and found there nothing of any value - no
> new messages around the time of the crash.
> 
> As this is a blade machine, most hardware issues (like memory problems)
> would be reported by the chassis, and it reports none.
> 
> I have a couple other M600 machines running OpenBSD, with the same
> network cards, and only once this problem did occur on another machine -
> a slave for the problematic one, running pfsync, sasync and acting as a
> carp backup.
> 
> Could anyone provide me with some tips on how to debug the problem
> further? Turn on some kernel debug mode, make use of the kernel
> debugger? I'm at my wits end here.
> 
> If there is any information that I could provide to help solve the
> problem, please don't hesitate to ask.
> 
> The dmesg is attached below:
> 
> <snip>

I'm currently testing a diff which disables hw vlan tag processing,
which I recieved off-list.

Aside from that, after increasing isakmpd verbosity, I started getting
messages like this before every vlan freeze:

> isakmpd[6940]: dropped message from 193.xxx.xxx.xxx port 500 due to
> notification type INVALID_FLAGS

Could they be somehow related?

-- 
regards, 

Wiesław Kielas

Reply via email to